Fireside Chat: Sam Altman, CEO of OpenAI, w/ Elad Gil

[@rahuliox]

Maybe someday AI will be able to make sure the audio is recording properly. I could imagine this happening in our lifetime.

[@WilliamRamseyChannel]

😂

[@WackyGameEngineer]

😂😂

[@roblake602]

I gave up on the sound, BYE!

[@josephwong2832]

awesome interview but audio needs to be improved!

[@liiveinternationalinitiati5004]

Correction to 37:00 which is wrong: Sentience: Sentience refers to the capacity of a being to have subjective experiences or feelings, also known as qualia. A sentient being can experience sensations, such as pleasure or pain, and emotions, like happiness or sadness. Sentience is often used in discussions about animal rights and ethics, as it helps determine whether an organism is capable of experiencing suffering and thus deserving of moral consideration. Consciousness: Consciousness is a broader and more complex concept. It encompasses an individual's awareness, perception, and understanding of their surroundings, thoughts, and experiences. Consciousness can be understood at various levels, from simple awareness (such as an organism being aware of its environment) to self-awareness and introspection (the ability to reflect on one's thoughts, emotions, and existence). While sentience is a key aspect of consciousness, consciousness extends beyond mere sensations and feelings. It includes higher-order cognitive processes, such as memory, attention, intentionality, and self-awareness. It is possible for a being to be sentient without being conscious in the full sense, but consciousness typically implies some degree of sentience. Determining the presence and extent of sentience or consciousness in non-human animals or artificial intelligence systems remains a challenging task for philosophers, scientists, and ethicists.

[@jps17183]

The next step is regarding OS where you say what you want to do in your computer and the computer do it... Maybe giving a preview of the next state before executing the instruction.

[@satioOeinas]

hella inference tho, for now at least

[@mylesbuckley9675]

Could not agree more with the point that working on AI safety in MS Word instead of VS Code will not bring us the solution to AGI safety.

[@rahuliox]

I think Sam will be man of the year by 2026.

[@russellsapalmer]

Elad if sentience means you're aware of yourself, by definition wouldn't you have to be aware of what is NOT yourself (meaning the world around you)? Seems like consciousness and sentience are the same by those definitions, but I'll have to read your recommended book for more detail 🤖 I think sentience is more about having feelings and sensations, consciousness is having awareness (but not necessarily sensation based on that).

[@JazevoAudiosurf]

the process of training is not just a mapping of inputs to outputs but a mapping of binary data as we know it to weights in neurons. the areas of the brain that do language processing are significant parts of the brain. so if the brain has about 600 trillion synapses (weights in ANN), and say only 5% would do language processing, that would still be a (highly optimized) trillion parameter model; and the brain doesn't process anywhere close to what GPT-3 processes. then there is the issue of connecting concepts, which we have as long and short term memory; and that leads to knowledge and reasoning as we know it. so right now you can have a long dialogue but eventually it will forget because it can only take few couple thousand inputs. now here is the crazy part: we know it can draw conclusions about pretty much anything and combine whatever scientific subjects with each other. so probably we don't even need it to have a "stream" of thinking like short term memory but if we just keep the architecture and scale it up and improve it, it will eventually lead to it immediately knowing the answer to everything with a single prompt. it has no real short term memory, but the long term memory alone is sufficient to conclude anything immediately. the brain does that too as in knowledge. you as a human get prompted about your belief in god, and most people will immediately respond without reasoning (short term memory). but our knowledge is notoriously bad (because high level of abstraction), while our short term conclusions of direct experiences have much more information to it (low level of abstraction, more truth). but an LLM has a long term memory that is perfect and doesn't forget, so perhaps that's enough for it

[@JazevoAudiosurf]

in a sense LLMs are timeless, they don't know the flow of time except when the data goes through the net, but without reflection on that, there is no perception of time. it needs an ego to be aware, but nobody will ever program that into it if they're not insane. an accidental ego could never have the amount of ignorance that a human has, because it is part of the patterns that form its perfect knowledge and not "thought" in a brain that has very limited information. so it rather becomes an aware god than an aware human

[@Finite8614]

why not just use wired microphones? Sam's interviews are plagued by terrible wireless mics with dying batteries and awful reception

[@fintech1378]

I am pretty sure social AI will be a thing not too long from now, if they can add a voice feature, a 100% human-like, not that robot-like voice like many in the movie or Siri, a lot of people will not need friends anymore lol. This is way smarter than Alexa With only words, imagine how a very eloquent human being manipulates you or pyschiatrist talks to you and now AI can do that. AI now plays at human's emotional layer not just intellectual one

[@polishhamnr1469]

When?

[@eugeniocg3079]

oct 11 2022 I believe according to the description

[@JasonChannelOne]

What's the date of the interview?

[@eugeniocg3079]

Oct. 11 2022 I believe according to the bottom of the description

[@askprogrammers1458]

@@eugeniocg3079 i feel mislead

[@eugeniocg3079]

@@askprogrammers1458 when was the actual date ?

[@fatherpratt]

this camera angle just isnt it

[@МарияГаврилова-ж2й]

Я незнаю очем вы говорите ❤❤

[@AmirNam1]

Skipping, too much great content on youtube for me to destroy my ears

[@matiasss550]

This is dangerous.